Feature: Medical
Figure 9: In the KWS20 case study, using higher clock frequencies with the RISC-V processor alone for loading and CNN management application resulted in lower energy consumption due to shorter loading and inference times. (Image source: Analog Devices)
consumption and time for loading model weights (kernels), loading input data and performing inference for typical edge AI applications. For example, in a case study of KWS with 20 keywords (KWS20), the results showed that developers could run the Arm processor alone to reduce loading time and energy consumption while running in diff erent MAX78000 power operation modes (Figure 8). T e study also examined the eff ect on
energy consumption and time when the Arm processor and RISC-V processor were asleep during idle time, with the RISC-V processor waking only long enough to perform loading and manage the CNN. Here, the study compared performance using two diff erent clock sources: T e MAX78000’s internal primary oscillator (IPO) at 100 MHz versus the lower power but slower internal secondary oscillator (ISO) at 60 MHz. In this result, a reduction in clock frequency dramatically
increased energy consumption associated with both loading and inference due to the longer completion time required for each (Figure 9). Based on their study, the Analog
Devices team noted that developers could achieve fast inference with minimal power consumption by running at higher clock rates, particularly with the high- performance Arm processor, employing judicious use of the MAX78000’s power operating modes, and retaining kernels in memory to avoid energy lost during extended loading times. For developers creating their own
edge AI solutions, Analog Devices off ers a comprehensive set of MAX78000 development resources, including its MAX78000EVKIT evaluation kit and MAX78000FTHR feather board. Along with an onboard digital microphone, motion sensors, colour display and multiple connection options, the
MAX78000EVKIT includes a power monitor feature to help developers optimise power consumption. For soſt ware development, Analog
Devices’ MAX78000 CNN toolset repository off ers documentation, development guides, training videos and soſt ware code supporting the evaluation kit and feather board. Building on an effi cient processor
subsystem foundation, Analog Devices presents a set of ultra-low-power microcontrollers that integrates the features and capabilities designed specifi cally to support the unique requirements of applications such as wearables, hearables, asset tracking, industrial and medical sensors and edge AI. Using these microcontrollers and supporting resources, developers can quickly implement designs that meet the specialised needs of diverse low-power applications.
www.electronicsworld.co.uk March 2025 37
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50
